Introduction:

Within this analysis, we’ll investigate factors correlated to diabetes. With a data set of 100,000 people, this investigation allows us to display relations between ages, HbA1c levels, smoking history, and glucose levels. With a wide range of data points, we begin to question if there are trends within this data that match our general understanding of diabetes. Our goal is to asses which of the 9 variables play a stronger role to the development of diabetes and if we can prove trends to better support our assumptions of this data. Through data visualization, chart analysis, and numerical analysis we will be able to present this data to convicne a general audience of the important factors that contribute to diabteic trends.

library(tidyverse) ## Loaded for dplyr
## Warning: package 'tidyverse' was built under R version 4.4.3
## Warning: package 'ggplot2' was built under R version 4.4.3
library(ggplot2) ## Loaded for plotting
library(plotly) ## Loaded for interactive plots
## Warning: package 'plotly' was built under R version 4.4.3
library(readr) ## Loaded to read in data
library(knitr) ## Loaded to compute and display data
library(scales) ## Loaded to scale data

Diabetes Dataset

100,000 × 9 (first 6 rows)
gender age hypertension heart_disease smoking_history bmi HbA1c_level blood_glucose_level diabetes
Female 80 0 1 never 25.19 6.6 140 0
Female 54 0 0 No Info 27.32 6.6 80 0
Male 28 0 0 never 27.32 5.7 158 0
Female 36 0 0 current 23.45 5.0 155 0
Male 76 1 1 current 20.14 4.8 155 0
Female 20 0 0 never 27.32 6.6 85 0

Male vs. Female Blood Sugar Levels (HbA1c) Plot

99,982 × 10 (first 6 rows)
gender age hypertension heart_disease smoking_history bmi HbA1c_level blood_glucose_level diabetes HbA1c_category
Female 80 0 1 never 25.19 6.6 140 0 Diabetes ≥ 6.5%
Female 54 0 0 No Info 27.32 6.6 80 0 Diabetes ≥ 6.5%
Male 28 0 0 never 27.32 5.7 158 0 Prediabetes 5.7% - 6.4%
Female 36 0 0 current 23.45 5.0 155 0 Normal < 5.7%
Male 76 1 1 current 20.14 4.8 155 0 Normal < 5.7%
Female 20 0 0 never 27.32 6.6 85 0 Diabetes ≥ 6.5%

Similar Prevalence of Prediabetes – The proportion of individuals categorized as having prediabetes (HbA1c 5.7% - 6.4%) is almost identical between males (41.3%) and females (41.4%). This suggests that prediabetes affects both genders at nearly the same rate.

Females Have a Slightly Higher Proportion of Normal Blood Sugar Levels – More females (38.4%) fall into the normal blood sugar category (<5.7%) compared to males (37.1%). This may indicate some slight protective factors or lifestyle differences in this group.

Since more males are in the diabetes category, there could be gender-related risk factors worth exploring—such as diet, activity levels, or genetic predisposition.

Overall, blood sugar regulation patterns appear fairly balanced between genders, but small differences suggest potential areas for further investigation.

Similar Prevalence of Prediabetes
The proportion of individuals classified as having prediabetes (HbA1c 5.7% - 6.4%) is nearly identical between males (41.3%) and females (41.4%). This suggests no significant disparity.

BMI Distribution by Hypertension Status Plot

Shows the distribution of BMI values based on hypertension status. A violin plot is great for visualizing the distribution and density of BMI across hypertension categories,

Shape and width: The width of each “violin” represents the density of BMI values at different levels. Wider sections mean more individuals have that BMI, while narrower sections indicate fewer people at those values.

Comparison of distributions: The blue violin represents people without hypertension (hypertension = 0), while the red violin represents those with hypertension (hypertension = 1). By comparing them, you can see how BMI differs between these groups.

The horizontal line around 25 BMI: This marks the median BMI for each group. Since both violins have a horizontal line in roughly the same position, it suggests that the median BMI is around 25 for both hypertensive and non-hypertensive individuals.

Density trends: If the violins have different thicknesses in certain BMI ranges, it tells you which BMI values are more or less common in each group. People with hypertension seem to have a higher BMI overall, but both groups share a similar median.

The distribution shape is different—for example, if one violin is wider at higher BMI values, it suggests that hypertension is more common among individuals with higher BMI.

Outliers or extreme values might appear as small bulges or extended tails at the ends of the violins, showing individuals with very high or low BMI.

The graph below is separated by whether or not a person has hypertension. With the comparison of BMI as the range, it’s seen that majority of people with and without hypertension lie within a BMI range of 25-29. Notice that for people with hypertension, the desnity population above the red line is greater than that of people without hypertension; indicating that there’s a larger of population of people with hypertension that have a larger BMI

library(ggplot2)
library(dplyr)

DHH_only <- diabetes_dataset %>% select(age, diabetes, heart_disease, hypertension) %>%  filter(age >= 2, diabetes == 1, heart_disease == 1, hypertension == 1)
No_DHH_only <- diabetes_dataset %>% select(age,heart_disease, diabetes, hypertension) %>% filter(age >= 2, diabetes == 0, heart_disease == 0, hypertension ==0)


ggplot() +
  geom_density(data = DHH_only, aes(x = age), fill = "#660033", alpha = 0.5) +  # Diabetes cases
  geom_density(data = No_DHH_only , aes(x = age), fill = "#66CCFF", alpha = 0.5) +
  labs(title = "Age Distribution: Diabetes vs. Heart Disease",
       x = "Age",
       y = "Density") +
  scale_x_continuous(breaks = seq(2, 80, by = 6)) +
  theme_bw() +
  theme(plot.title = element_text(hjust = 0.5)) +
  theme(
    plot.title = element_text(size = 13),
    axis.text = element_text(size = 11),
  )

BMI vs. Age Across Diabetes & Heart Disease Plot

A relation to BMI and Heart Disease

Each person within this scale has heart disease. Here a comparison is made between declared underweight and overweight people, grouped by sex, based on a BMI scale. There’s a significant increase in population percentage for those who are considered overweight and that have heart disease. With visual aid, it can be concluded that as weight increases, chances of heart disease will increase.

Smokers go brrr# Smokers go brrrgeom_boxplot()

In the smoking data there are 6 unique values

  1. Never: Has Never smoked
  2. Not current: Has smoked but is not currently smoking
  3. Former: Has quit smoking (abstained for longer than)
  4. Current: Is currently a smoker
  5. Ever: Has ever smoked regardless of current smoking status
  6. No Info: No smoking history information available

The total amount of people who fall into each category is as follows;

  1. Never: 35095
  2. Not current: 6447
  3. Former: 9352
  4. Current: 9286
  5. Ever: 4004
  6. No Info: 35816

There is quite a sizable amount of people in the No info category.

The total number of people in the dataset is 100000. To help clean up the data, we can filter ‘No Info’ people out. When we do that we get 64184.

# Figure out the unique categories of smoking history
unique(diabetes_dataset$smoking_history)
## [1] "never"       "No Info"     "current"     "former"      "ever"       
## [6] "not current"
# Count amount of people who belong to each unique smoking category
#Omit No info

smoking_tally <- diabetes_dataset %>% filter(smoking_history != 'No Info') %>%  group_by(smoking_history) %>% summarise(total_people = n())

#group diabetic vs non diabetic people together

smoking_diabetes_dataset <- diabetes_dataset %>%
  filter(smoking_history != 'No Info') %>%  
  group_by(smoking_history, diabetes) %>%  
  summarise(total = n())
## `summarise()` has grouped output by 'smoking_history'. You can override using
## the `.groups` argument.
smoking_diabetes_dataset
## # A tibble: 10 × 3
## # Groups:   smoking_history [5]
##    smoking_history diabetes total
##    <chr>              <dbl> <int>
##  1 current                0  8338
##  2 current                1   948
##  3 ever                   0  3532
##  4 ever                   1   472
##  5 former                 0  7762
##  6 former                 1  1590
##  7 never                  0 31749
##  8 never                  1  3346
##  9 not current            0  5757
## 10 not current            1   690
# Inner join tally data with diabetic grouped data;
#mutate a column to create a percentage per category;
#select desired columns

smoking_diabetes_percentage <- inner_join(smoking_tally, smoking_diabetes_dataset, by = 'smoking_history') %>% mutate(Percentage = total/total_people *100) %>% select(smoking_history, diabetes, total, Percentage)

smoking_diabetes_percentage
## # A tibble: 10 × 4
##    smoking_history diabetes total Percentage
##    <chr>              <dbl> <int>      <dbl>
##  1 current                0  8338      89.8 
##  2 current                1   948      10.2 
##  3 ever                   0  3532      88.2 
##  4 ever                   1   472      11.8 
##  5 former                 0  7762      83.0 
##  6 former                 1  1590      17.0 
##  7 never                  0 31749      90.5 
##  8 never                  1  3346       9.53
##  9 not current            0  5757      89.3 
## 10 not current            1   690      10.7

Now we can graph the relationship between smoking and diabetes as separated by smoking category.

library(ggplot2)
library(plotly)

#Create initial graph about smoking and diabetes

smoking_diabetes_graph <- ggplot(smoking_diabetes_percentage) + 
  geom_col(aes(x = smoking_history, y = total, fill = as.factor(diabetes)), position = 'dodge') 

smoking_diabetes_graph <- smoking_diabetes_graph + coord_flip() + labs(
  title = 'Smoking History and Diabetes Relationship',
  y = 'People',
  x = 'Smoking History',
  fill = 'Has Diabetes',
  caption = "1 indicates diabetes, 0 indicates no diabetes"
  ) 

smoking_diabetes_graph

ggplot(smoking_diabetes_percentage) + geom_col(aes(x = smoking_history, y = Percentage, fill = as.factor(diabetes)))